Resea ch A icle
Vib o ac ile S imula ion as an Ins uc o o Mimic y-Based
Physical Exe cise
Jani Lylykangas,1Jani Heikkinen,2Veikko Su akka,1Roope Raisamo,1
Kalle Myllymaa,1and A o Lai inen3
1Tampe e Uni o Compu e -Human In e ac ion (TAUCHI), School o In o ma ion Sciences, Uni e si y o Tampe e,
Kansle in inne 1, 33014 Tampe e, Finland
2The Uni o Human-Cen e ed Technology (IHTE), Depa men o Pe asi e Compu ing, Tampe e Uni e si y o Technology,
P.O. Box 553, 33101 Tampe e, Finland
3Audio Ride s, L d., En allinkuja 4B, 01900 Nu mij¨
a i, Finland
Co espondence should be add essed o Jani Lylykangas; jani.lylykangas@u a. i
Recei ed 18 June 2015; Re ised 11 Sep embe 2015; Accep ed 29 Sep embe 2015
Academic Edi o : Ma co Mamei
Copy igh © 2015 Jani Lylykangas e al. This is an open access a icle dis ibu ed unde he C ea i e Commons A ibu ion License,
which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed.
The p esen aim was o in es iga e unc ionali y o ib o ac ile s imula ion in mimic y-based beha io al egula ion du ing physical
exe cise. Vib o ac ile s imuli communica ed ins uc ions om an ins uc o o an exe cise o pe o m lowe ex emi y mo emen s.
A wi eless p o o ype was es ed i s in con olled labo a o y condi ions (S udy 1) and was ollowed by a use s udy (S udy 2) ha
was conduc ed in a g oup exe cise si ua ion o elde ly pa icipan s wi h a new e sion o he sys em wi h imp o ed cons uc ion
and ex ended unc ionali y. The esul s o S udy 1 showed ha ib o ac ile ins uc ions we e success ul in bo h supplemen ing and
subs i u ing isual knee li ins uc ions. Vib o ac ile s imuli we e accu a ely ecognized, and exe cise wi h he de ice ecei ed
a i ma i e a ings. In e es ingly, ac ile s imula ion appea ed o s abilize accele a ion magni ude o he knee li s in compa ison o
isual ins uc ions. In S udy 2 i was ound ha use expe ience o he sys em was mainly posi i e by bo h he exe cise s and hei
ins uc o s. Fo example, exe cise wi h ib o ac ile ins uc ions was expe ienced as mo e mo i a ing han con en ional exe cise
session. Toge he he esul s indica e ha ac ile ins uc ions could inc ease possibili ies o people ha ing di icul ies in ollowing
isual and audi o y ins uc ions o ake pa in mimic y-based g oup aining. Bo h s udies also e ealed de elopmen a eas ha
we e p ima ily ela ed o a sligh delay in igge ing he ib o ac ile s imula ion.
1. In oduc ion
S uc u ed physical g oup exe cise is ypically based on seeing
and lis ening o ins uc ions o a aine and hen modeling
hem by ainees who y o do he same mo emen s in
synch ony wi h he aine . The e a e many g oups, howe e ,
likeagedanddisabledpeoplewhocanno bene i o ge he
ull u ili y om such ac i i ies i hei senso y o cogni i e
impai men s inhibi hem om ollowing isual and audi o y
ins uc ions. U ilizing he sense o ouch as an in o ma ion
channel could pa ly eplace he equi emen s o he ull
unc ionali y o isual and audi o y pe cep ion (see [1] o
a e iew). Following his, ouch media ed ins uc ions could
inc ease he achie abili y o g oup exe cise o special g oups
and i could ac ually complemen he exe cise e en in he case
o ully unc ional senso y sys ems.
Mo o esponse o a ouch sensa ion is an inna e ai o
human beings, and he sense o ouch has some ad an ages
o e he o he senses [2]. Fo example, eac ion imes (RTs)
o ac ile s imuli ha e been ound o be as e han RTs o
isual s imuli (e.g., [3]). As compa ed o audi o y RTs, ac ile
RTs ha e been ound o be less a ec ed by dis ac ions such
as a concu en seconda y ask (e.g., [4]). This is p obably
because, in con as o isual and audi o y e en s, ac ile
e en s a e in di ec physical con ac wi h he o ganism and
hus e oke mo e eadily he in e es o an o ganism [5].
Mechanisms o eac o ac ile s imula ion a e among he
i s ones o de elop as e idenced by he palma g asp e lex
Hindawi Publishing Co po a ion
Ad ances in Human-Compu e In e ac ion
Volume 2015, A icle ID 953794, 13 pages
h p://dx.doi.o g/10.1155/2015/953794
2Ad ances in Human-Compu e In e ac ion
o e uses [6]. Fu he mo e, he impo ance o he sense
o ouch has been shown by, o example, emaining o
mo o esponses o ac ile s imula ion, while he pa ien s a e
un esponsi e o o he ypes o s imula ions like isual and
audi o y s imula ions [7].
Rela i ely ecen expe imen al esea ch has shown ha
echnology-media ed ib o ac ile s imula ion (i.e., cu aneous
s imula ion p oduced by a ib a ion ac ua o ) can be success-
ullyusedinassis ingphysical ehabili a ionandexe ciseby
means o senso y subs i u ion and augmen a ion. Wea able
ib o ac ile eedbacksys emsha ep o en obeusable, o
example, in gai e aining [8], body balance con ol [9],
owing [10], snowboa ding [11], and e en swimming [12].
The e is also e idence ha app op ia ely designed ib o ac ile
signals can be in ui i ely associa ed wi h cues sugges ing,
o example, o accele a e and decele a e he speed o mo e-
men [13]. The e idence on in ui i e associa ions be ween
ac ile s imula ion and human beha io sugges s ha ac ile
s imula ion migh also unc ion as a subs i u e o ano he
pe son’s isual demons a ion o bodily mo emen s. Lee
and his colleagues [14, 15] ha e empi ically es ed body
balance aining wi h ac ile in o ma ion in he mimic y
ame o e e ence, bu o he wise using ac ile s imula ion
in p o iding in e pe sonal, mimic y-based beha io al egu-
la ion in o ma ion seems o be s ill la gely uns udied. In a
p elimina y s udy by Lee and Sienko [14] he ask was o
eplica e p e eco ded unk il mo emen s consis ing o a
20∘an e io il ollowed by a 6 s long s a ic hold and a 20∘
pos e io unk il back o he up igh s a ing posi ion. The
il mo emen s we e pe o med a a a e o abou 1.12∘/s. The
ins uc ions we e ei he isual, ac ile, o combined isual-
ac ile cues. Visual ins uc ion was p o ided wi h a side- iew
illus a ion o a i ual a a a ep esen ing he p e eco ded
example mo emen s o an ins uc o . Tac ile ins uc ions
we e p o ided wi h wo sepa a e ib a ion ac ua o s placed
on unk midline in he na el and back indica ing he an e io
and pos e io il s, espec i ely. End o he ib a ion commu-
nica ed he a ainmen o he desi ed il angle. The esul s
showed ha he example il s we e mimicked signi ican ly
mo e accu a ely and as e wi h combined isual- ac ile and
ac ile ins uc ions as compa ed o isual ins uc ions. In a
ollow-up s udy [15] he speed o he example mo emen was
a ied, and his ime only ac ile ins uc ions we e p o ided.
The esul s showed ha slow (app oxima ely 1.12∘/s) example
mo emen s we e mimicked mo e accu a ely as compa ed
o medium (app oxima ely 2.0∘/s) and as (app oxima ely
4.0∘/s) example mo emen s. Toge he he esul s o Lee and
colleagues [14, 15] indica e a se ious po en ial o using ac ile
ins uc ions o mimic physical mo emen o ano he pe son
in eyes- and hands- ee ashion. Howe e , mo e esea ch is
needed o assess he unc ionali y o he concep in o he
ypes o mimic y-based physical aining.
In con as o ea lie wo k, ou aim was o b oaden
he scope om pe sonal (i.e., one- o-one) balance ain-
ing owa ds as e -paced hy hmic knee li exe cise o en
pe o med in g oup aining si ua ions. Clea ly, he cha -
ac e is ics and equi emen s o he ela i ely slow-paced
unk il exe cise pe o med in he ea lie s udies and he
hy hmic knee li exe cise (e.g., ma ching in place) wi h
many pa icipan s de ia e om each o he in many espec s.
Fo example, he p ecise pos u al mimic y o he angula
magni ude o he ins uc o ’s mo emen conside ed essen ial
in he ea lie unk il s udies was no ecognized as a
cen al ac o in he hy hmic knee li ing ask. This is because
com o able li heigh ange may be widely a ied wi hin
he exe cise g oup depending on he pa icipan s’ physical
condi ion. Accu a e empo al mimic y enabling hy hmic
synch ony, howe e , is essen ial in his ype o an exe cise in
o de okeep heg oupin hesamephase.So,incon as
o he ins uc ions gi en in balance aining, i is mo e i al
ocommunica e hes a ime a he hanendpoin o he
mo emen . Ano he change conside ed necessa y o he
cu en in es iga ion was he eplacemen o he i ual a a a
used in [14] by a human ins uc o o mee he acili ies
o a ypical g oup exe cise si ua ion. Impo an ly, we also
add essed he exe cise s’ subjec i e expe iences on how hey
conside ed ouch in o ma ion in he con ex o physical
exe cise.
We designed a sys em ha measu ed an ins uc o ’s leg
mo emen s and, on he basis o hese measu emen s, ac ile
s imula ions we e wi elessly ou ed o a leg o a pa icipan
o ins uc he momen s o pe o m leg li s. Two e sions o
a wi eless and wea able p o o ype sys em we e designed in
an i e a i e ashion o in es iga ion. In S udy 1, he i s p o-
o ype e sion was es ed in con olled labo a o y condi ions
wi h indi idual adul s pa icipa ing in a simula ed exe cise
si ua ion o ge basic in o ma ion on he unc ionali y o he
sys em and o e eal main de elopmen a eas. In S udy 2,
hesecondp o o ype e sionwi hmodi iedcons uc ionand
inc eased unc ionali y needed in ac ual aining si ua ion
was es ed in ealis ic condi ions wi h a po en ial a ge g oup
consis ing o ge ia ic pa icipan s aking pa in a s uc u ed
g oup exe cise. In bo h s udies he pa icipan s pe o med
mimic y-based asks wi h isual and ac ile guidance and
hen e alua ed hei expe iences o pe o ming he exe cise.
2. S udy 1: Labo a o y Expe imen
2.1. In oduc ion. The aim o he i s expe imen was o
e alua e how a wi eless ac ile guidance applica ion unc ions
in compa ison o con en ionally used isual demons a ion
in a knee li exe cise. Typically in such exe cise an ins uc o
shows how and when a knee should be li ed, and he
exe cise ’s ask is o mimic he demons a ion mo emen s in
he same phase wi h he ins uc o .
We used p e eco ded ideo ins uc ions in gi ing he
isual guidance s imuli o p o ide consis en and compa able
es condi ions. A e lec ion symme y mo o pa adigm by
Ba elas e al. [16] was used o ins uc he pa icipan o
imi a e he ins uc o ’s demons a ion mo emen s shown
in he ideo so ha he pa icipan and he ins uc o
we e acing each o he and he ask was o espond o he
ins uc o ’skneeli asi wa chingone’sownimage e lec ed
by a mi o .
Tac ile guidance was p o ided wi h ib a ion s imula ion
on a leg. Tac ile s imulus was based on a push me apho
[11, 17]. This means ha he ask was o wi hd aw he leg
Ad ances in Human-Compu e In e ac ion 3
away om he s imulus loca ed in he do sal side o he leg
abo e c ook as hough he ib a ion would “push” he limb
o an opposi e di ec ion. The s imulus- esponse compa ibili y
be ween push e sus pull s imula ion and consecu i e body
mo emen s has been s udied o some ex en (e.g., o w is
o a ion di ec ion s udied in Jansen e al. [17]), bu so a
es ablished guidelines o op imal s imulus loca ions, o
example, o a knee li mo emen , seem o be nonexis en .
Loca ion, du a ion, ib a ion equency, and ampli ude o
he ac ile s imula ion we e chosen based on a pilo es
whe e six pa icipan s e alua ed ib a o y s imula ions in
di e en leg loca ions while pe o ming a knee li exe cise.
The o e all p e e ence o he wo me apho s was almos
equally dis ibu ed and hus he decision was based on he
sligh ly mo e a i ma i e s imulus ampli ude a ings gi en o
he push me apho s imula ions.
The e we e h ee modali y condi ions o p o ide knee li
ins uc ions: isual (V), ac ile (T), and isual- ac ile (VT)
guidance. Knowing ha a sligh delay be ween he onse s o
isual and ac ile s imuli will ine i ably be p esen due o an
accele a ion h eshold-based igge ing o he ac ile s imulus
(i.e., he ins uc o ’s leg mo emen needed o exceed a ce ain
magni ude o se on he ac ile s imulus in he exe cise ’s leg)
as well as ope a ing la ency o he ib o ac ile ac ua o , we
ocusedoncompa ing hekneeli RTbe ween hemodali y
condi ions. RT was conside ed as an impo an measu e
because synch onous-enough iming be ween he onse s o
ins uc o ’s demons a ion mo emen and he ac ile guid-
ance s imulus would be a p e equisi e o a success ul g oup
exe cise especially wi h isually impai ed exe cise s. We also
in es iga ed possible e ec s o ac ile s imula ion on he leg’s
accele a ion magni ude o s udy he mo emen dynamics
be ween he di e en ins uc ion modali ies in mo e de ail.
In addi ion o he objec i e measu es o he knee li RT and
accele a ion, he pa icipan s’ subjec i e expe iences we e
inqui ed o assess how hey a ed he ac ile ins uc ions in
his ype o ac i i y.
2.2. Me hods
2.2.1. Pa icipan s. Twel e olun a y pa icipan s (3 emale
and 9 male) be ween he ages o 19 and 48 yea s (M =26
and SD =9) ook pa in he s udy a e signing in o med
consen o ms. They we e s uden s o s a o he Uni e si y
o Tampe e. All epo ed ha ing no mal sense o ouch and
no mal o co ec ed o no mal ision. Ele en o hem we e
igh leggedandonewasle leggedby hei own epo .The
s udy was app o ed by a local esea ch e hics commi ee.
2.2.2. S imuli. The knee li guidance s imuli we e p esen ed
in h ee modali y condi ions. In V modali y condi ion he
s imuli we e p o ided wi h an exe cise ideo (320 × 240
pixel esolu ion, ∼30 Hz sampling a e) whe e an ins uc o
acing owa ds he pa icipan p oduced he s imuli by
demons a ing le knee li mo emen s (Figu e 1).
In VT modali y condi ion, a 200 ms long and 285 Hz
ib a ion s imulus in he pa icipan ’s igh high was coupled
wi h each knee li e en p esen ed in he exe cise ideo.
Figu e 1: A sc eensho o he exe cise ideo showing he ins uc o
pe o ming a demons a ion mo emen .
The cu en s imulus du a ion has been ea lie used in
wea able ib a ion sys ems o body mo emen ins uc ions
by, o example, Sienko e al. [9] and Jansen e al. [17]. The
ib a ion equency was wi hin he bes sensi i i y ange o
Pacinian co puscle mechano ecep o s a ound 10–500 Hz [18]
and was ound o be well pe cei ed in he p eexpe imen al
pilo es . The ib a ion was p o ided in an ea ly phase o he
demons a ion knee li mo emen as soon as he ins uc o ’s
leg accele a ion exceeded a iaxial sum ec o h eshold o
0.75 g calcula ed as
√(𝑔𝑋2)+(𝑔𝑌2)+(𝑔𝑍2)(1)
in a baseline-co ec ed da a. The baseline was he mean o
he i s 200 samples om he beginning o he measu emen
while he ins uc o was s anding s ill. The h eshold was
de e mined by inspec ing accele a ions o he demons a ion
mo emen s shown in he exe cise ideos. The da a showed
ha he cu en h eshold was equi ed o be exceeded in
o de o eliably de ec he upwa d phase o each o he
demons a ion mo emen s. In consequence o he accel-
e a ion h eshold he e was an a e age o 203 ms (SD =
31ms) delay be ween he ins uc o ’s ini ial leg mo emen
and he onse o he ac ile s imulus. In T modali y con-
di ion he ib a ion s imuli we e iden ical o hose in VT,
bu he isual ins uc ions (i.e., he exe cise ideo) we e
excluded. In o de o educe he p edic abili y o s imulus
onse ime, in e s imulus in e al (ISI) was a ied andomly
wi hin h ee du a ions: 3 (sho ), 6 (medium), and 9 (long)
seconds, app oxima ely. Th ee sepa a e exe cise ideos (i.e.,
one o each modali y condi ion) wi h di e en ly a anged
ISIoccu encewe eused oelimina epossiblelea ninge ec
ega ding he unning o de o he sho , medium, and long
ISI ials.
2.2.3. Appa a us. The i s p o o ype de ice (Figu e 2) was
buil on A duino Duemilano e p o o yping boa d ha was
used o p ocessing he da a om an accele ome e , con ol-
ling a ib a ion mo o and communica ing wi elessly wi h a
PC o e Blue oo h. The p o o yping boa d, Blue oo h mod-
ule, and 9 V ba e y used o powe he de ice we e enclosed
o a box and placed in o a bel bag. P ecision Mic od i es
4Ad ances in Human-Compu e In e ac ion
(a) (b)
Figu e 2: (a) Wea able pa s o he appa a us including elec onics in he enclosu e box (middle), accele ome e inside he ound housing
( igh ), and ib a ion mo o inside he ube housing (le ). Scale is indica ed by a coin in diame e o 20 mm. (b) The bel bag con aining he
enclosu e box and he leg band equipped wi h he accele ome e and ib a ion mo o wo n by a pa icipan .
ib a ion mo o (model 304-111) encapsula ed in a plas ic
ube was used o s imula ing a leg. The ypical ope a ing
cha ac e is ics o he ib a ion mo o included a lag ime o
16 ms and a ise ime o 28 ms. A 3.3 V ou pu o he A duino
boa d was used oge he wi h a bipola ansis o o d i e he
ib a ion mo o . The ansis o was con olled o e he digi al
ou pu o he p o o yping boa d. Dimension Enginee ing DE-
ACCM3D iaxial accele ome e senso encapsula ed in a
plas ic housing measu ed he accele a ion o a pa icipan ’s
leg mo emen . The accele ome e ea u ed a ±3gsense ange
and 500 Hz bandwid h o 𝑥-axis, 𝑦-axis, and 𝑧-axis. An
analog inpu o he p o o yping boa d was used o ead he
ou pu o he accele ome e . The ib a ion mo o and he
accele ome e we e a ached o a leg band. The PC and he
p o o ype de ice communica ed wi h he BlueSMiRF Sil e
module o e se ial po . A USB-Blue oo h adap e was used
o ansla ing he wi eless communica ion in a sampling a e
o 60 Hz. The PC was used o sending he ac ile ins uc ion
s imula ion and p esen ing he ideo ins uc ions o he pa -
icipan and o s o ing he ecei ed accele a ion da a om
he pa icipan ’s leg using cus om so wa e p og ammed in
P ocessing.o g language. The so wa e was used o ecei ing
accele a ion da a o e he wi eless communica ion om he
p o o ype de ice and o s o ing he ime s amped da a
o a ile o la e analysis. The ins uc ions we e based on
p e eco ded accele a ion measu emen s ha we e eco ded
wi h he de ice and s o ed oge he wi h he eco ded exe cise
ideo in synch oniza ion. Visual ins uc ions we e p esen ed
h ough a 24󸀠󸀠 Samsung SyncMas e display.
2.2.4. P ocedu e. A e wea ing he bel bag including he
elec onics, helegbandwas as eneda ound hepa icipan ’s
igh high so ha he accele ome e was loca ed in he en al
sideabou 10cmabo e hekneecapand he ib a ionmo o
was in he do sal side o he high (Figu e 2(b)). Then a
hea ing p o ec o was pu on and wo 200 ms long es s imuli
we e p esen ed o con i m ha he pa icipan was able o
eel he ib a ionandunable ohea he ib a ionsound.
The pa icipan was guided o abou 120 cm dis ance om
he compu e display o which cen e poin was 155 cm abo e
he loo . Then he expe imen e ga e e bal ins uc ion and
a isual example o he exe cise mo emen . The ask was o
li he igh knee un il he high angle was pa allel wi h he
loo , a he maximum, and immedia ely a e ha o lay he
leg down back in he s a ing posi ion. The li was ins uc ed
o be done as soon as he isual ins uc ion on he ideo
and/o ac ile s imulus was pe cei ed. This ins uc ion was
gi en once be o e he ac ual expe imen al ials. Running
o de o V, T, and VT condi ions and he h ee exe cise ideos
was ully coun e balanced be ween he pa icipan s. A simila
p ocedu e was ollowed in V, T, and VT condi ions apa om
he ollowing a ibu es:
(i) In V condi ion, he ideo was u ned on and he
ib a ion mo o o .
(ii) In T condi ion, he ideo was u ned o and he
ib a ion mo o on.
(iii) In VT condi ion, bo h he ib a ion mo o and he
ideo we e u ned on.
A e he expe imen e had le he oom, each condi ion
s a edwi hanini ia ion ialexcluded om heanalysisand
was ollowed by 21 expe imen al ials consis ing o se en
sho , medium, and long ISI ials. An expe imen al ial
s a ed wi h a andomly chosen sho , medium, o long ISI
which was ollowed by V, T, o VT ins uc ion s imulus
depending on he modali y condi ion. The pa icipan ’s ask
was oimmedia ely espond o heins uc ions imulus
byakneeli and hens a owai o anex ial.
One condi ion ook abou 2.5 minu es. A e comple ing a
condi ion, he expe imen e e u ned o he es ing oom
and he pa icipan was allowed o ha e a sho b eak. Then
he expe imen e announced he ins uc ion modali y o he
upcoming condi ion. A e pe o ming all he h ee modali y
condi ions he pa icipan assessed hem wi h a ing scales. In
acco dance wi h Meh abian and Russell [19] and B adley and
Lang [20] he scales we e bipola nine-poin scales a ying
om −4 o+4.Ra ingswe easked o hepleasan nesso
he ask ( a ying om unpleasan o pleasan ), easiness o
he ask ( a ying om di icul o easy), and in e es o he
ask ( a ying om bo ing o in e es ing). On each scale,
0 ep esen ed a neu al expe ience (e.g., nei he unpleasan
no pleasan ). Ra ings we e gi en using pen and pape .
Ad ances in Human-Compu e In e ac ion 5
Finally, he pa icipan was asked o a e he ac ile s imula-
ion ega ding i s ampli ude ( oo weak, app op ia e, o oo
s ong) and du a ion ( oo sho , app op ia e, o oo long). In
addi ion, he empo al uni o mi y o he iming o he ac ile
s imula ions in ela ion o he isual ins uc ions ( oo ea ly,
app op ia e, o oo la e) p esen ed in he VT condi ion was
asked o alida e hesubjec i eexpe ienceo helagcaused
by he echnical implemen a ion o he ac ile s imulus.
2.2.5. Da a Analysis. RTs and peak accele a ions o he knee
li s we e analyzed wi h a one-way epea ed measu es analysis
o a iance (ANOVA) wi h modali y condi ion as a wi hin-
subjec s ac o . I he sphe ici y assump ion o he da a was
iola ed, G eenhouse-Geisse co ec ed deg ees o eedom
we e used o alida e 𝐹s a is ic. Bon e oni co ec ed pai -
wise 𝑡- es s we e used o pos hoc es s. RTs we e measu ed
om he momen when he ins uc o ’s demons a ion knee
li exceeded he 0.75 g accele a ion h eshold o he momen
when he pa icipan ’s knee li mo emen eached i s peak
accele a ion. Peak accele a ion sum ec o was de ined as he
highes alue du ing an upwa d phase o he pa icipan ’s
knee li e en using a simila baseline de ini ion and o mula
as in de e mining he h eshold o s imulus onse in he
exe cise ideo. A o al o 33% o he RT and accele a ion
da a (i.e., 4, 5, and 3 pa icipan s in V, T, and VT condi ions,
esp.) was excluded due o echnical easons causing noisy
accele a ion signal and eplaced by a condi ion speci ic se ies
means in he analyses. P io o he impu a ion, he no mali y
o he ob ained RT and peak accele a ion da a was con i med
using Shapi o-Wilk es o no mali y. The es s sugges ed ha
he assump ion o no mal dis ibu ion was me in he V, T,
and VT condi ions ega ding RT (S-W = 0.95, d = 8, and
𝑝 = 0.67;S-W=0.95,d =7,and𝑝 = 0.76;S-W=0.97,d =9,
and 𝑝 = 0.91, esp.)andpeakaccele a ionda a(S-W=0.90,
d = 8, and 𝑝 = 0.28;S-W=0.95,d =7,and𝑝 = 0.72;S-W=
0.87, d = 9, and 𝑝 = 0.13, esp.).
Ra ings o pleasan ness, easiness, and in e es we e ana-
lyzed wi h h ee sepa a e F iedman es s. I he F iedman es
showed a signi ican e ec , Wilcoxon Signed-Rank es s we e
used o pai wise compa isons.
2.3. Resul s. Means and s anda d e o o he means (SEMs)
o heRTsandpeakaccele a ionso kneeli sa ep esen ed
in Figu e 3. Nobody missed any knee li s imuli in he V, T,
and VT condi ions.
2.3.1. Reac ion Times. A one-way epea ed measu es analysis
o a iance (ANOVA) showed a s a is ically signi ican e ec
o modali y: 𝐹(1.27,14.00) = 15.66;𝑝 ≤ 0.001.Pos hoc
pai wise compa isons be ween he modali ies showed ha
he RTs we e signi ican ly as e in V and VT condi ions as
compa ed o T condi ion: MD = 158 ms and 𝑝 < 0.05 and
MD = 183ms and 𝑝 < 0.001, espec i ely. RT di e ence
be ween V and VT condi ions was s a is ically nonsigni ican .
2.3.2. Peak Accele a ions. Aone-wayANOVAshoweda
s a is ically signi ican e ec o modali y: 𝐹(2,22) = 5.91;𝑝<
0.01. Pos hoc pai wise compa isons be ween he modali ies
VTVT
Ins uc ion modali y
Reac ion ime
Peak accele a ion
0
200
400
600
800
Time (ms)
0
1
2
3
4
Accele a ion (g)
Figu e 3: Mean RTs and mean peak accele a ions o knee li s in
isual (V), ac ile (T), and isual- ac ile (VT) modali y condi ions.
E o ba s ep esen SEMs.
Pleasan ness Easiness In e es
V
T
VT
−4
−3
−2
−1
0
1
2
3
4
Mean a ing
Figu e 4: Mean a ings o pleasan ness, easiness, and in e es o
he knee li asks in isual (V), ac ile (T), and isual- ac ile (VT)
modali y condi ions. E o ba s ep esen SEMs.
showed ha he peak leg accele a ions we e signi ican ly
g ea e in VT condi ion when compa ed o T condi ion: MD
=0.25gand𝑝 ≤ 0.05. O he pai wise compa isons we e
s a is ically nonsigni ican .
2.3.3. Subjec i e Ra ings. Means and SEMs o he subjec-
i e a ings a e shown in Figu e 4. F iedman es showed
a s a is ically signi ican e ec o he a ings o in e es :
𝜒2= 9.66;𝑝 < 0.01. Fo he a ings o pleasan ness and
easiness, he F iedman es e ealed no s a is ically signi ican
e ec s. Wilcoxon Signed-Rank es s showed ha he ask was
conside ed in T and VT condi ions as signi ican ly mo e
in e es ing han in V condi ion: 𝑍 = −2.51and 𝑝 < 0.05and
𝑍 = −2.45 and 𝑝 < 0.05, espec i ely. Di e ence be ween T
and VT condi ions was no s a is ically signi ican .
Vib a ion ampli ude was conside ed as app op ia e by
each o he wel e pa icipan s. No one epo ed he ib a ion
being oo weak o oo s ong. Vib a ion du a ion was assessed
as app op ia e by ele en pa icipan s, while one conside ed

6Ad ances in Human-Compu e In e ac ion
ha he ib a ion was oo long. Nobody conside ed he
ib a ion oo sho . In VT condi ion he ib a ion iming
(i.e., synch ony) wi h espec o he knee li shown in
he ins uc ion ideo was ag eed as app op ia e by i e
pa icipan s and he emainde conside ed ha he ib a ion
came oola e.Nobody hough ha he ib a ionwas oo
ea ly wi h espec o he ideo.
2.4. Discussion. The esul s o S udy 1 showed ha ac ile
ins uc ion channel was unc ional in a knee li exe cise
bo h alone (i.e., T condi ion) and when combined wi h
isual ins uc ions (i.e., VT condi ion). I seems ha he
pa ame e s (i.e., du a ion, ib a ion equency, ampli ude,
and loca ion) o he ac ile s imulus we e applicable because
nobody missed any ac ile s imuli. In addi ion, he indings
in he subjec i e a ings and pos expe imen al in e iews
showed a o able assessmen s o ac ile ins uc ions. In he
in e iews s imulus ampli ude and du a ion we e conside ed
as app op ia e almos unanimously. The empo al synch ony
o he isual and ac ile ins uc ion in VT condi ion, howe e ,
pola ized he opinions, as mo e han hal o he pa icipan s
hough ha he ib a ion came oo la e wi h espec o knee
li shown in he ideo.
The expe ienced asynch ony was likely due o he ela-
i ely obus accele a ion h eshold causing he 203 ms delay
and he 16 ms onse la ency o he ib a ion ac ua o , which
oge he esul ed in a minimum o abou 220 ms delay om
he ins uc o ’s ini ial leg mo emen o he onse o he
ac iles imulusin hepa icipan ’sleg.Inconsequence, he
ins uc o ’s leg mo emen was al eady isible in he ideo
when he ac ile s imulus s a ed. The delay can also explain
why he mean RT (i.e., ime o each he peak leg accele a ion)
in T condi ion was 158.2 and 182.5 ms slowe as compa ed o
V and VT condi ions, espec i ely.
I he ac ile s imulus onse had been c ea ed in exac
synch ony wi h he ins uc o ’s ini ial leg mo emen , he RTs
in T and VT condi ions could, in ac , ha e been abou 40–
60 ms as e han hose in V condi ion. This conclusion is
based on sub ac ion o he lag caused by he echnical imple-
men a ion om he mean RT o T condi ion and suppo ed
by ea lie indings showing abou 30–40 ms sho e RTs o
ac ile han isual s imuli (e.g., [3, 5, 21]). Following his,
ac ile ins uc ions would seem o ha e he po en ial o enable
e en as e knee li eac ions han con en ionally used isual
ins uc ion p o ided ha he onse o he ac ile s imulus
couldbe eliablysi ua ed o heini ialphaseo heins uc o ’s
leg mo emen . Howe e , he aim o he p esen s udy was
o assess he unc ionali y o ac ile ins uc ion sys em in
a ealis ic exe cise si ua ion, whe e a obus -enough onse
me hod would be equi ed o p e en he ins uc o om
gi ing alse-posi i e ins uc ion s imuli esul ing, o exam-
ple, om mino o unin en ional leg mo emen s be ween
he epe i ions. Thus, lowe ing he accele a ion h eshold o
diminish he onse delay would equi e u he s udies o ind
an accep able ade-o be ween he speed and accu acy o
p o iding he ac ile ins uc ions.
Ins uc ionmodali yalsohadane ec o heaccele a ion
magni ude o he leg mo emen s. VT ins uc ion esul ed
in signi ican ly highe peak accele a ion han T ins uc ion.
Despi e he ac ha he mean peak accele a ion was he
highes in V condi ion and he lowes in T condi ion, he
di e ence did no each s a is ical signi icance. As i can be
seen in he SEM e o ba s o Figu e 3, his was due o he
la ge de ia ion o accele a ion magni ude in V condi ion
compa ed o hose ound in T and VT condi ions. This
esul may indica e ha ac ile s imula ion s abilized he
accele a ion o leg mo emen and enabled mo e consis en
epe i ions.
Ra ings o he pleasan ness, easiness, and in e es o he
exe cise ask we e clea ly posi i e in each modali y condi ion
which indica es ha ac ile ins uc ion channel was well
accep ed in he cu en con ex along wi h con en ionally
used isual ins uc ions bo h unimodally and when com-
binedwi h isualins uc ionmodali y.Thiswasdespi e he
ac ha hepa icipan swe e ullycapableo ollowing isual
ins uc ions and, he e o e, did no necessa ily need ac ile
ins uc ions in his ype o ac i i y. Mo eo e , ac ile s imula-
ion p o ided in T and VT condi ions inc eased signi ican ly
he pa icipan s’ in e es o he ask when compa ed o V
condi ion. This inding sugges s ha such a de ice may be
o in e es o exe cise s also in eal aining si ua ions, which
in u n would be impo an in he ligh o ea lie s udies
showing ha in e es owa ds he aining is a cen al ac o in
inc easing he mo i a ion and exe cise adhe ence (e.g., [22]).
3. S udy 2: Use Expe ience and Accep ance
o a Vib o ac ile Exe cise De ice among
Elde ly Pa icipan s
3.1. In oduc ion. In a ollow-up use s udy, he wea able
ac ile de ice used in S udy 1 was i e a ed u he and
edesigned obeusedinag oupexe cisesi ua ionwi h
elde ly pa icipan s. This app oach was chosen based on a
ecen end ha has been shi ing inc easingly om one-
o-one pe sonal aining o g oup exe cise and aining
p og ams [23]. Supe ised g oup exe cise is o en a o ed in
public and ins i u ional heal hca e, because i is clea ly mo e
ime-e ec i e and cos -e ec i e in compa ison o pe sonal
aining and i has been ound o be bene icial in p omo ing
no only physical i ness [24] bu also social and men al well-
being and o e all quali y o li e [25].
Anew e siono hede iceallowedanins uc o o
ac i a e an exe cise ’s legs wi h ac ile s imula ion au o-
ma ically and in eal ime (i.e., in conjunc ion wi h he
ins uc o ’s demons a ion mo emen s) ia a wi eless link.
The main goal o his use s udy was o e alua e accep ance
and use expe ience (UX) o he ac ile exe cise de ice in
a con ex o ec ea ional g oup exe cise among a po en ial
a ge popula ion.Theapp oachhad h eemain esea ch
ques ions:
(1) How do elde ly pa icipan s expe ience he use o a
ac ilede iceinag oupexe cisecon ex ?
(2) How does i di e om con en ional session wi hou
ac ile ins uc ions?
Ad ances in Human-Compu e In e ac ion 7
(a) (b)
Figu e 5: (a) Wea able pa s o he exe cise de ice and (b) hei a achmen loca ions.
(3) Do he elde ly pa icipan s accep he use o ac ile
s imula ion in a g oup exe cise con ex ?
The use s udy consis ed o exe cise sessions bo h wi h
and wi hou he de ice, UX ques ionnai es, and in e iews
inqui ing he exe cise s’ as well as he ins uc o s’ imp es-
sions, expe ience, and accep ance o using he de ice.
3.2. Me hods
3.2.1. Pa icipan s. Pa icipan s and p emises o he s udy
we e p o ided by a se ice cen e o elde ly people. The
a ge ed use g oup consis ed o ela i ely heal hy elde ly
olun ee s wi hou any majo cogni i e disabili ies o ensu e
ha hey we e ully able o answe ques ionnai es and ake
pa in an in e iew. They we e also equi ed o be able o
ake pa in he physical exe cise asks, including knee li ing
used in S udy 1. Two male and six emale exe cise s and
wo emale ins uc o s ook pa in he s udy a e signing
in o med consen o ms. The exe cise s we e be ween he
ages o 73 and 92 yea s (M =83and SD =7). The ins uc o s
we e 27 and 39 yea s old. One o he ins uc o s had a o mal
educa ion in physical he apy. The s udy consis ed o wo
wi hin-subjec s exe cise g oups, each wi h one ins uc o and
ou exe cise s. The s udy was app o ed by he local esea ch
e hics commi ee.
3.2.2. Appa a us. The ac ile de ice consis ed o wo elas ic
leg bands and a bel bag (Figu e 5). Bo h he leg bands
included a ac ile ac ua o and accele ome e , while he bel
bag con ained an A duino boa d and a 9 V ba e y. The
componen s in he leg bands we e connec ed o he boa d
ia la cables. The s udy was done wi h wo iden ical de ices.
Onede icewo nbyanins uc o unc ionedina“mas e ”
mode and he o he wo n by an exe cise was se in a “sla e”
mode. The de ices communica ed ia Blue oo h wi h a PC
ac ing as hos and ou ing he communica ion be ween he
de ices. A mo e de ailed desc ip ion o he sys em’s ope a ion
p inciple is gi en in [26]. The majo change om he use s’
iew poin was he ac ha , con a y o he de ice used in
S udy 1, leg bands we e a ached o bo h limbs. In addi ion,
he ib a ion ac ua o ype was eplaced wi h Sola bo ics
VPM2 Vib a ing Disk Mo o due o i s closed s uc u e.
The e was a mechanical delay o 60 ms, on a e age, measu ed
om heonse o heapplied ol age(3V) o hebeginning
o he ac ua ion om he mo o using a GW Ins ek GDS-
2104A oscilloscope. The accele ome e s we e eloca ed abo e
he ac ua o s in he do sal side o he legs. Toge he hese
e isions enabled a mo e compac design o he leg bands.
3.2.3. Use Expe ience Ques ionnai e. The UX ques ionnai e
was used o s udying he esea ch ques ions 1 and 2.
The ques ionnai e had wo a ian s: one o e alua ing he
exe cise expe ience while wea ing he de ice and ano he
wi hou i . They had nea ly iden ical scales and dis inc i e
ins uc ions on wha o e alua e. The ques ionnai es we e
based on UX model by Hassenzahl [27]. The seman ic
di e en ials we e selec ed om wo di e en ques ionnai es:
A akDi 2 [28] and A ak-Wo k [29]. A akDi 2 ques-
ionnai e consis s o se en-poin seman ic di e en ial scales,
which a e di ided in o p agma ic quali ies (PQ), hedonic
quali ies o s imula ion (HQS) and iden i ica ion (HQI),
and a ac i eness (ATT). The A ak-Wo k adds a ask
and goal achie emen (TGW) quali y, which is essen ial
in wo k- ela ed con ex s. The p agma ic scales a e close o
he adi ional de ini ion o usabili y (use ulness, u ili y, and
usabili y). I a p oduc is s imula ing, i p o ides possibili ies
o gain new knowledge and sel -de elopmen , while iden i-
ica ion can be explained as social sel -exp ession h ough
physical objec s and possessions. Th ee de ice-cen ic scales
we e omi ed om he ques ionnai e wi hou he de ice:
imp ac ical, p ac ical; makes exe cise ha de , makes exe cise
easie ; and slows down he exe cise, speeds up he exe cise.
A ew scales we e selec ed o each o he i e hemes o keep
he ques ionnai e leng h on one page. Mo eo e , some scales
had obele ou as heywe ep esumed obeinapplicable
o his s udy due o he con ex and na u e o he de ice. The
selec ed scales o each quali y a e p esen ed in Table 1.
3.2.4. In e iew. The g oup in e iews concen a ed on qual-
i a i e in o ma ion on he pa icipan s’ gene al imp essions
o ac ile ins uc ions and he de ice. The in e iews we e
semis uc u ed and s a ed wi h ee- o m desc ip ion o he
pa icipan s’ imp essions o he use expe ience. This was
ollowed by a discussion abou he exe cise wi h and wi hou
8Ad ances in Human-Compu e In e ac ion
Table 1: Quali ies and scales used in he UX ques ionnai es.
Quali y Scale ID
P agma ic
Complex, simple PQ 2
Imp ac ical, p ac ical PQ 3
Challenging, e o less PQ 4
Con using, clea PQ 6
Hedonic:
s imula ion
Con en ional, o iginal HQS 1
Unimagina i e, c ea i e HQS 2
Dull, abso bing HQS 5
Hedonic:
iden i ica ion
Unp o essional, p o essional HQI 2
Poo quali y, high quali y HQI 4
Noninclusi e-Inclusi e HQI 5
Sepa a es me om people, b ings me
close o people HQI 6
A ac i eness
Unpleasan , pleasan ATT 1
Una ac i e, a ac i e ATT 3
Bad, good ATT 5
Discou aging, mo i a ing ATT 7
Task and goal
achie emen
Makes exe cise ha de , makes exe cise
easie TGW 1
Slows down he exe cise, speeds up he
exe cise TGW 4
Di icul ocon ol,easy ocon ol TGW 7
he de ice, accep ance, and o he ideas o he de ice and
applica ions bo h he exe cise s and ins uc o s migh ha e.
3.2.5. P ocedu e. The exe cise s’ ask was o ollow and epea
heexe ciseguidedby heins uc o .Theexe ciseconsis ed
o mo ing legs in sea ed and s anding posi ions, including
a ious leg li ing asks, s a iona y ma ching, and dancing.
The whole exe cise was done wi h backg ound music. The
sessions we e designed and ca ied ou by he ins uc o s
as no mally as possible bu wi h emphasizing on legs and
lowe pa o he body. The exe cise sessions began wi h an
in oduc ion o hepu poseo heuse s udyandab ie
explana ion o he de ice. The exe cise s we e old ha he
ins uc o ’s de ice sen ac ile ins uc ions o he exe cise ’s
de ice as an addi ional signal o s a pe o ming he exe cise
mo emen s. Tac ile s imuli we e p oduced concu en ly in
bo h he ins uc o ’s de ice and he exe cise ’s de ice, hus
enabling he ins uc o o con i m a success ul sending o a
ac ile ins uc ion. Due o he limi ed amoun o de ices, he
exe cise ’s de ice was o a ed be ween he ou exe cise s in
he g oup. E e y 15 minu es he exe cise was paused o swi ch
he de ice o a nex exe cise . As a esul , each exe cise go
o wea and exe cise wi h he de ice o abou 15 minu es
du ing he1-hou session.Theins uc o wo e heo he
de ice o he whole exe cise session. Hal way h ough he
session, he exe cise s illed he UX ques ionnai es e alua ing
he subjec i e expe ience o exe cising in he g oup. The i s
wo exe cise s o y he ac ile de ice illed he ques ionnai e
“exe cise wi h he de ice,” and he o he wo illed he ques-
ionnai e “exe cise wi hou he de ice.” This was epea ed a
heendo hesessionbu his ime ice e sa.A heend,
each exe cise in he g oup e alua ed he exe cise wi h and
wi hou wea ing he de ice. The ins uc o s in bo h g oups
e alua ed he use expe ience o exe cising wi h he de ice
a e he session. Following he session, he g oup (including
he ins uc o ) ook pa in 30-minu e-long in e iew.
3.3. Resul s
3.3.1. Use Expe ience Ques ionnai e. The exe cise s’ a ings
gi en o he UX ques ionnai e (Figu e 6(a)) showed ha
he exe cise bo h wi hou and wi h he de ice was expe i-
enced mainly in a posi i e way. They e alua ed he exe cise
wi h he de ice o be mo e mo i a ing han ha wi hou
he de ice (ATT 7), bu o he wise he SEM anges o he
a ings we e o e lapping be ween he wo exe cise si ua ions
indica ing po en ially nonsigni ican di e ences. Ra ings o
he h ee i ems e lec ing he de ice use si ua ion only we e
also posi i e; he exe cise s conside ed ha he de ice was
mo e p ac ical han imp ac ical (PQ 3), made he exe cise
somewha easie (TGW 1), and speeded up he exe cise
(TGW 4). In e es ingly, he ins uc o s’ a ings o he de ice
in Figu e 6(b) we e mainly ou side he SEM ange o hose
gi en by he exe cise s shown in Figu e 6(a). One o bo h
o he ins uc o s a ed he de ice mo e posi i ely han he
exe cise s ega ding 8 i ems (i.e., PQ 2, PQ 6, HQS 1, HQS 2,
HQI 2, HQI 5, HQI 6, and ATT 7) and mo e nega i ely
ega ding 10 i ems (i.e., PQ 3, PQ 4, HQI 4, ATT 1, ATT 3,
ATT 5, ATT 7, TGW 1, TGW 4, and TGW 7). Consensus
was ound only o i em HQS 5; he exe cise wi h he
de ice was ound equally abso bing by he exe cise s and
he wo ins uc o s. Howe e , no s a is ical analysis was
pe o med o he ques ionnai e da a due o low numbe o
he pa icipan s.
3.3.2. In e iew: Exe cise Expe ience wi h and wi hou he
De ice. Exe cise s’ expec a ions o he exe cise wi h he
de ice we e summa ized in o h ee g oups: (1) no expec-
a ions, (2) sligh ly ne ous o a aid, and (3) in e es ed.
Majo i y o he exe cise s we e sligh ly ne ous o a aid
o ying a de ice hey did no ha e any expe ience wi h.
Howe e , a e he exe cise s ied he de ice, many o he
expec a ions we e changed in a posi i e way as exempli ied
by ollowing quo a ions:
“I was ne ous abou his hing we we e supposed
o wea . Then I saw how simple i was and he ea
anished,” an exe cise said.
“A i s I was ne ous and su p ised. I didn’
expec i o be his easy,” an exe cise said.
On he o he hand, one exe cise commen ed ha his/he
expec a ions did no change, and ano he one had expec ed
much s onge s imula ion.
Fi s imp essions o exe cise wi h he ac ile de ice, o
bo h he exe cise s and ins uc o s, we e posi i e and e en
su p ising. These consis ed o posi i e and neu al desc ip-
ions, such as nice, pleasan , easy, un, o dina y, di e en (i.e.,
Ad ances in Human-Compu e In e ac ion 9
Simple
P ac ical
E o less
Clea
O iginal
C ea i e
Abso bing
P o essional
High quali y
Inclusi e
B ings me close o people
Pleasan
A ac i e
Good
Mo i a ing
Makes exe cise easie
Speeds up exe cise
Easy o con ol
[PQ_2] Complex
[PQ_4] Challenging
[PQ_6] Con using
[HQS_1] Con en ional
[HQS_2] Unimagina i e
[HQS_5] Dull
[HQI_2] Unp o essional
[HQI_4] Poo quali y
[HQI_5] Noninclusi e
[HQI_6] Sepa a es me om people
[ATT_1] Unpleasan
[ATT_3] Una ac i e
[ATT_5] Bad
[ATT_7] Discou aging
[TGW_7] Di icul o con ol
∗)[PQ_3] Imp ac ical
∗)[TGW_1] Makes exe cise ha de
∗)[TGW_4] Slows down exe cise
Exe cise s wi hou de ice
Exe cise s wi h de ice
−2 −1 0123
−3
Mean a ing
(a)
Simple
P ac ical
E o less
Clea
O iginal
C ea i e
Abso bing
P o essional
High quali y
Inclusi e
B ings me close o people
Pleasan
A ac i e
Good
Mo i a ing
Makes exe cise easie
Speeds up exe cise
Easy o con ol
Complex
Imp ac ical
Challenging
Con using
Con en ional
Dull
Unimagina i e
Unp o essional
Poo quali y
Noninclusi e
Sepa a es me om people
Unpleasan
Una ac i e
Bad
Discou aging
Makes exe cise ha de
Slows down exe cise
Di icul o con ol
Ins uc o 1 wi h de ice
Ins uc o 2 wi h de ice
1023
−2 −1−3
Mean a ing
(b)
Figu e 6: Mean a ings o he UX ques ionnai es on 7-poin seman ic di e en ial scale gi en by (a) he exe cise s and (b) he ins uc o s.
I ems labeled wi h ∗)in (a) we e excluded om he “exe cise wi hou he de ice” ques ionnai e. E o ba s in (a) ep esen SEMs.
some hing new), and e o less. The emphasis on he easiness
was equen in he commen s. Once he de ice was pu on,
he exe cise s did no need o concen a e on ope a ing he
de ice. The majo i y o he exe cise s also quickly unde s ood
he pu pose o he de ice wi h a e y sho in oduc ion. Two
o he exe cise s needed u he ins uc ions. One exe cise
had majo di icul ies in ecognizing he ac ile s imula ion
and ano he was so concen a ing on ollowing he ac ile
ins uc ions ha she o go o pay a en ion o he ac ual
exe cise. The ac ile s imuli and he o m- ac o o he wea -
able de ice we e com o able and pleasan acco ding o he
exe cise s. Howe e , one o he ins uc o s said ha wea ing
he de ice o an hou became uncom o able because o he
equen ac ile s imula ion.